1 Flashcards
(166 cards)
1
Q
What do direct and indirect age standardisation use
A
Direct = population indirect = death rates
2
Q
Advantages of indirect age standardisation
A
Useful if small numbers in group
data that is available
3
Q
How do you write relative risk of 1.43
A
1.43 times as likely to develop outcome with exposure
4
Q
How do you write attributable risk
A
x number of cases among exposure can be attributable to exposure
5
Q
how you write the attributable fraction
A
x% of cases among exposure can be attributable to exposure
6
Q
What is Healthy screenee bias
A
volunteers more likely to be fit and healthy
7
Q
Disadvantages of case control studies
A
Poor for rare exposures
Bias - eg recall
Problem of selection of controls
cannot estimate risk of disease in exposed / unexposed
8
Q
what are the criteria of causality
A
temporal dose response strength of association biological plausibility consistency reversibility specificity coherence analogy
9
Q
If an association is not causal what could it be
not a true association
A
Bias
Chance
confounding
reverse causality
10
Q
What is primary prevention ? Eg?
A
Measures to prevent healthy disease free people from developing disease
Immunisations to prevent infectious diseases
11
Q
What is secondary prevention ?
Eg
A
Prevent individuals with asymptomatic disease from developing symptoms
Breast cancer screening
12
Q
What is tertiary prevention?
Eg?
A
Measures to prevent symptomatic patients developing complications
Diabetic retinopathy screening
13
Q
Focus Criteria for screening
A
Condition
method
treatment
programme
14
Q
Screening criteria condition
A
Should be an important problem
natural history should be understood
recognisable early / latent period
15
Q
Screening criteria method
A
There should be a suitable method for detection
should be acceptable to the population
16
Q
screening criteria treatment
A
needs to be an accepted treatment
17
Q
screening criteria programme
A
policy on who to treat
facilities for diagnosis should be available
costs of detection should be balanced to overall spending
18
Q
purpose of randomisation
A
minimise selection bias
equally distribute known / unknown confounders
19
Q
how to work out and Interpret sensitivity
A
a/a+c
x% of screened population that have disease will test positive
20
Q
How to work out specificity and interpret ?
A
d/b+d
x% of the disease free will test negative
21
Q
How to work out negative predicted value ? interpret
A
d/c+d
x% of those with a negative result will be disease free
22
Q
How can you improve PPV
A
PVs are dependent on prevalence of disease. Selecting a higher risk group will improve the PPV
(NVP will be lower risk group)
23
Q
How to work out PPV
A
A/A+B
number of people with a positive test who actually have disease
24
Q
What is issue with low response rate
A
non responders could be different from responders in important ways
especially if non response is related to exposure and outcome
25
interpret odds ratio of 1.8
80% higher odds of *outcome* with *exposure*
26
If confidence intervals include 1 what does it mean
could be due to chance
27
2 towns
A - more smoking, more asthma
B - less of each
Why cant you say smoking causes more asthma
Ecological fallacy - Data from population cant be assumed for individual
Cant control confounders
Migration of populations - alter results
data on exposure and outcome may be collected in different ways
28
What is epidemiology
Study of frequency, distribution, and determinants of diseases in populations to prevent and control disease
29
what is secondary attack rate ? how to work out?
number of new cases among contacts
| number of cases in contacts / number of contacts
30
What is sufficient / component / necessary cause?
S- inevitably produce outcome
c - one of the factors
N- needed for outcome
31
difference between odds and risk
odds - comparison of exposure and who didn't
| risk - exposure vs everyone
32
How to control confounding at study design
restriction
randomisation
matching
33
Control confounding in analysis
Stratification
| Statistical modelling using multivariable regression
34
What is stratification
Measures association between each outcome separately for each category of confounder
35
What does having residual confounders do? what type of misclassification is this?
biases exposure and outcome in same direction as the confounder
non-differential random misclassification
36
what is bias
Systematic deviation between comparison groups wich may misinterpret the association being investigated
37
What is validity
Accuracy - the degree to which a measurement measures what is proposes to measure
38
what is reliability
the degree to which results can be replicated
39
Egs of information bias ?
| Sources?
inacurate measurement / classification of exposure / outcome
observer (recall), participant, instument
40
What is differential misclassification?
Usual cause ?
direction of bias
2 populaitons measured in different ways / with different instuments
Observer / responder bias
Either way
41
What is no differential misclassification ?
| what does it result in?
both comparison groups are equally likely to be misclassified
Underestimation of true association
42
What happens to relative risk with non-differential misclassification
moves closer to 1
43
how to avoid bias
Random selection
blinding
Automated instruments
use objective (rather than subjective) measures eg. records rather than recall
44
What 2 things are considered for sample size calculation? explain them both
statistical power - probability of determining if effect is real
precision - probability of detecting an effect if it is not real (by chance)
45
In analysis what does random sampling enable
use of interferal statistics - allows generalisation
46
Why would you not always use simple random sampling
Hard to obtain sampling frame
population spread over large area
want to include minorities
47
What is stratified sampling ? benefits
each member assigned to a group then simple random of each group
include minorities
48
What is cluster
members assigned to a cluster - sampling then of all in a cluster
49
egs of indirect data collection
medical records, census, registries
50
how can you reduce errors in data collection tools
use a pilot study
51
Types of data variable
binomial - eg yes/no
categorical
continuous - eg blood pressure
52
what is a regression model
way of describing relationship between variable and potential explanatory factors
Eg Oucome and exposure / outcome and confounding
53
What is an effect modifier ? eg?
measure of association changes with 3rd variable
| Risk of heart disease higher in heavy smokers vs lighter smokers and then greater at older age (age is effect modifier)
54
Advantages / disadvantages of using population strategy in prevention
a- Large potential for population, behaviorally appropriate
d - Small benefits to individuals, poor motivation of subject, benefit to risk ratio may be low.
Inequality within population - who more likely to adhere
55
Advantages / dis of using high-risk strategy for prevention
A- Appropriate to individuals, subject motivated, benefit to risk ratio favourable
D- Screening costs, temporary / limited effect, may not seek to reduce 'risky' behaviour, can cause ANXIETY
56
Levels of health promotion
Individual - encouraging smokers to stop through education and support
Population - banning smoking in public areas
Screening
57
What is the prevention paradox
If high-risk individuals only represent a small proportion of population then risk reduction might NOT have a population-level reduction.
When compared to a slight risk reduction at population level
58
What does screening need to be for an individual
Inexpensive, easy to administer, impose minimal discomfort
59
When can screening be primary prevention
screening to identify risk factors
60
Who can be screened in a program?
Mass- whole pop
Targeted - groups at risk
Opportunistic - eg during clinic visits
Systematic - eg smear test every 4 years
61
what is criteria for screening called
Wilson and Jugner criteria
62
Ethical issues in screening
benefits must outweigh risks
False positives -> unnecessary anxiety
False negatives -> False security and failure to recognise subsequent symptoms
may need counselling
63
Egs of bias in screening
Selection - people who participate may be different
lead time
length time
64
what ratio can be calculated from cross sectional
prevelance ratio
65
what ratio can be calculated from ecological studies
risk ratio, odds ratio, incidence ratio
66
what ratio can be calculated from cohort studies
risk ratio, odds ratio, incidence ratio
67
what ratio can be calculated from case-control studies
odds ratio of exposure
68
what ratio can be calculated from intervention studies
risk ratio, odds ratio, incidence ratio
69
what type of study compares groups at a point in time
ecological
70
what study examines groups over time
time trend study
71
what does a time trend study investigate
if changes in incidence vary with exposures over time
72
what could cause time trend to not work
if the outcome has a long latent period
73
benefits of studying populations
see group level effects
public health interventions aimed at groups
availability of data
cheap and quick
74
benefits of multi group study
compare various exposures / outcomes
Good to compare spatial patterns in frequency of outcome
group level exposure likely to be more accurate than individual level measures
can generate hypothesis
Find out what factors affect population level and think how to modify
75
what is vital registration data
births and death by cause
76
what is used to classify disease? issue with this
Internaltional classification of disease
| risk of information bias related to accuracy of diagnosis
77
sources of data in ecological ? eg of individual level data ? group level?
demographic, vital registration, disease registries
Individual - cancer incidence / mortality rates
Group - %of skilled birth attendants / life expectancy
78
2 methods of analysis for ecological
correlation - how closely 2 variables linked
| regression - mathematically works out to enable prediction of one variable to another
79
issues with interpretation of ecological
```
cant show causation
unable to adjust for confounders
bias - in data collection
ecological fallacy
migration of populations - mixing data
```
80
what is ecological fallacy
cannot assume group level associations also apply at the individual level
81
when can migration of populations be usefull
determining genetic vs environment causes
82
pros of cross sectional
```
quick and easy to perform
provides prevalence of risk factors
useful when planning health services
repeated studies can monitor change over time
No loss to follow up
```
83
There is no loss to follow up in cross sectional but what can happen
people refuse to answer questions eg. sexual activity
84
what are descriptive cross sectional studies
look at frequency and distribution of exposures / outcome at a point / period (eg. 2 weeks)
85
what are analytical cross sectional
look at association between exposure to risk factors and outcome at same time
86
bias in cross sectional
recall - may not remember if exposure was a long time ago / more likely to remember if they are aware of link
non response - are they different to responders
representativeness (selection bias)
87
Issues with cross sectional
Can't show causality - just association
bias
only studies prevalent cases
exposure and disease collected simultaneously
88
What is issue with only studying prevelant cases in cross sectional
less likely to include short duration cases
89
Why is exposure and disease collected simultaneously an issue ? eg
exposures may change over time.
Eg diet may change to reduce discomfort in colon cancer
Can lead to REVERSE CAUSALITY
90
How to highlight potential issues with study
pilot study
91
what is decided first in cohort - exposure or outcome
exposure
92
2 types of cohort
prospective
| retrospective
93
Pros of prospective
doesnt need records
| observations to see what outcomes develop
94
pros of retorspective
Useful for investigating long term hazzards
95
3 parts of cohort
1- selection of population
2- exposure assessment
3- follow up and outcomes
96
if you have a common exposure how do you select study population? rare exposure? what other population fo you need?
common - slect population before classifying on exposure
rare - select on bias of exposures
Unexposed
97
what can be issue with exposure assessment
can change over time - eg smoking status
98
when to calculate risk ratio ve rate ratio
follow up times are similar - risk
| different - rate
99
what can be done if exposures change over time?
time series analysis
100
bias in cohort
```
selection bias (including loss to follow up)
information bias
observer bias - classification of exposure / outcome
```
101
what can using 2 observers lead to
systematic misclassification bias
102
pros of cohort
useful for rare exposure
can study multiple outcomes
can assess dose response (and see if there is a threshold)
meet temporality criteria for causality
103
limitations of cohort
need a large sample size
expensive and time
loss to collow up
Retrospective - may lack data on confounders / innacurate data on exposure
104
Advantages of case control studies
```
relatively quick / cheap
can study mulitple exposures
can be used to study rare outcomes
Useful when rapid result is required
Better than cohort when long latency
Useful for Genetic epidemiology
Low loss to follow up
```
105
When might a rapid result be required
Outbreak investigations
106
Disadvantages of case control
Exposure and outcome already occured - susceptible to selection / information bias / recall bias
Potential for reverse causality
Not suitable for rare exposures / multiple outcomes
No estimate of incidence / prevalence
107
How can you reduce confounding in case control ? Issue with this ?
Matching
| Can be complicated / costly
108
what is a nested case-control study ? What does this enable
Cases are members of cohort with outcome and controls are members without
Automatic matching
Avoids selection / information bias and reverse causality
109
What do you need to try and do with sampling for case-control
As patients are selected on bias of outcome need to make sure they are representative of target population
Need a case definition and whether to include both prevalent and incident cases
110
Benefits of including prevalent cases in case control?
Make it easier to generalise to target population
111
Issues of using prevalent cases? Cc
May differ from incident cases and reduce validity of study
May lead to underepresentation of more severe cases (who die sooner after diagnosis and so less likely to be selcted for study) or less severe (who recover)
Reverse causality - exposure factors changing
112
Issue with using prevalent cases for chronic conditions
Make it hard to determining exposure to risk factors that vary over time -> lead to reverse causality (EG diet and colon cancer)
113
If outcome is fatal in case control what needs to be done
Include patients who have dies to avoid selection bias
114
What needs to be considered in rare/ unusual cases for case control
those sent out of catchment area for specialist treatment - should be included
115
What is over matching ? what is issue?
Use too many similar characteristics or select characteristics too closely associated with exposure
Cases and controls dont differ enough to main exposure - Can't measure association
116
What is frequency matching
eg by gender
| 60% of cases female use 60% female controls
117
How to avoid observer bias when collecting retrospective data
Researcher doesnt know if they are researching case or control
118
Bias in cases (case control)?
Recall bias
119
What is generated in analysis of case control? what cant be ?
Odds ratio of exposure
| Can't directly estimate incidence / prevelance or frequency of exposure in general population
120
If cases are matched what is used for analysis ? CC
Conditional logistic regression
121
What do you need to look out for in interpreation of case control?
Prone to selection / recall bias
| Need to avoid observer bias and reverse causality
122
What is the difference between an efficacy and and effectiveness trial
Efficacy - done under ideal conditions - check maximum effect
Effectiveness - determine effect in routine clinical conditions
123
How do you calculate Number needed to treat
Use absolute risk reduction
124
Advantages of intervention studies
Infer causality - satisfy temporality consideration
If removal of exposure - satisfy reverse causality
Randomise - enables distribution of all known / unknown confounders
Can blind - reduce information bias
125
Disadvantages of intervention studies
Not approriate for rare outcomes
Long and expensive
If intervention is already in use - unsafe / hard to withdraw
Unethical if intervention is shown to be effective (for those on control)
126
What are the WHO guidelines for studies based on
the declaration of Helsinki
127
What needs to be attained before any study
Ethical approval
128
What is a plausibility study
Evaluates effectiveness of interventions by comparison with a control group but WITHOUT randomization
129
When are plausability studies useful?
Intervention so complex RCT will be too artificial
When intervention has shown effectiveness on small scale but large scale needs to be demonstrated
Ethical concerns prevent RCT
130
3 types of control in plausability studies
Historical
Geographical
Opportunistic
131
How does a historical control work ? Issue with therse
Compare frequency of outcome in group before and after intervention was used.
Can't distinguish from other changes which may have occured
132
Benefit of Geographical controls
Adjust for temporal effects
133
Type of study design with geographical controls ? How does it work
Steeped wedge design
| Slowly add intervention to different areas
134
When are opportunistic controls used? What else can happen with them?
People who should have had intervention but didnt (Eg if program didnt reach them)
Recieve varying degrees of intervention - assess dose resposne
135
4 phases of RCT
1 - check saftey and tolerability in healthy
2- saftey and efficcacy in larger groups (maybe RCT)
3- provide evidence in efficacy at risk of outcome in RCT
4- monitor routine use of intervention without a comparisson group
136
When are cluster RCTs useful
When intervention affects groups (eg pollution)
| Risk of contamination between intervention and control (eg info leaflets)
137
What is it called when 2 or more interventions compared individually and combined against control? Why are these useful?
Factoral design
| Allows interactions to be assessed
138
What is a crossover design RCT ? Benefit? When cant they be used?
Individuals act as own control with a washout period between control and intervention
Adjusts for all confounders
Needs to have intervention which doesnt have ling term effects
139
Why might you get selection bias in RCTS
Volunteers may be more or less likely to take part with some characteristics
People who drop out may be different
140
What is minimisation in intervention studies? what is issue?
Purposely allocate individuals
| Risk of selection bias (Eg more affected individuals)
141
How to avoid bias in intervention
blinding
142
How to avoid bias in analysis of intervention ?
Use a statistician independent of study twam for intervention allocation
143
When do you need interim analysis?
Long follow up period with possibility of severe adverse effects / sufficient evidence intervention is working
144
Difference between intention to treat and per-protocol?
ITT - compares all participants and includes drop outs
| PP - only includes participants who take the drug properly to see true potential
145
why should you include drop outs? Other benefit of ITT
Intervention may make disease worse and excluding drop outs would make intervention appear more effective
Closer to what would be seen in real world - missed doses / on time / normal use
146
Why is per protocol useful?
If you include non proper taking eg. late vaccines you can get non differential missclassification bias - underestimate potential of drug
147
How do you calculate intervention efficacy
1-relative risk
148
How to calculate NNT
Absolute risk reduction (Incidence in control - incidence in intervention) Get answer Eg 4 cases /1000
NNT = 1000/4 = 250 cases to prevent one case of disease
149
How to calculate and interpret AR
Incidence in exposed - incidence in unexposed
| X number of cases of exposed can be attributed to exposure
150
How to calculate AF ? Interpret?
AR/incidence in exposed
| X% of cases in exposed can be attributed to exposure
151
Alternate calculation of AF
(Relative risk-1) / relative risk
152
How to calculate preventable fraction ? 2 ways
(Incidence in unexposed - incidence in exposed ) / incidence in unexposed
1-relative risk
153
How to calculate PAR ? Interpret
Incidence in population - incidence in unexposed
Incidence of outcome in population can be attributed to exposure
154
What is issue with PAR
Assumes causality
155
How to calculate PAF ? Interpret?
PAR / incidence in pop
| The proportion of cases in the population that are attributable to the exposure
156
Alternate way of calculated PAF
E(relative risk -1) / ( E(relative risk -1) +1)
E= exposed population
157
What is age an example of in relation to smoking and CHD?
| Increased age increases risk alongside increased smoking
Effect modifier
158
What is calculated at end of direct age standardisation ? indirect?
d - age standardised rate ratio
| i - Smr
159
How do you calculate DSR
Expected / total pop - gives answer as a rate
| compare rate to that of original pop
160
Why is case-control bad for temporality criteria
exposure and outcome at same time
| Potential for reverse causality / exposure changing over time
161
What is bias
error that results in a systematic deviation from the true estimation of the association between exposure and outcome
162
define Selection bias
a systematic error in
- the selection of study participants or
in the allocation of participants to different study groups
163
When does information bias occur
is inaccurate measurement or classification of
- exposure or
- outcome.
Information bias occurs when there is differential misclassification, which occurs when there is a systematic error in measurement or classification
164
What type of bias is healthy worker effect
selection
165
What is healthy worker effect and what does it cause
Workers tend to be healthier than population
| If you compare workers to population may result in underestimation of excess risk due to exposure
166
Advantages of indirect over direct standardisation
More intuitive
Data available
Useful if numbers in strata are small
